Article

Adaptive Modifications of Muscle Phenotype in High-AltitudeDeer Mice Are Associated with Evolved Changes in GeneRegulationGraham R Scott1 Todd S Elogio1 Mikaela A Lui1 Jay F Storz2 and Zachary A Cheviron3

1Department of Biology McMaster University Hamilton ON Canada2School of Biological Sciences University of Nebraska Lincoln3Department of Animal Biology University of Illinois Urbana-Champaign

Corresponding author E-mail scottg2mcmasterca chevironillinoisedu

Associate editor Michael Nachman

Abstract

At high-altitude small mammals are faced with the energetic challenge of sustaining thermogenesis and aerobic exercisein spite of the reduced O2 availability Under conditions of hypoxic cold stress metabolic demands of shivering ther-mogenesis and locomotion may require enhancements in the oxidative capacity and O2 diffusion capacity of skeletalmuscle to compensate for the diminished tissue O2 supply We used common-garden experiments involving highland andlowland deer mice (Peromyscus maniculatus) to investigate the transcriptional underpinnings of genetically based pop-ulation differences and plasticity in muscle phenotype We tested highland and lowland mice that were sampled in theirnative environments as well as lab-raised F1 progeny of wild-caught mice Experiments revealed that highland natives hadconsistently greater oxidative fiber density and capillarity in the gastrocnemius muscle RNA sequencing analyses re-vealed population differences in transcript abundance for 68 genes that clustered into two discrete transcriptionalmodules and a large suite of transcripts (589 genes) with plastic expression patterns that clustered into five modulesThe expression of two transcriptional modules was correlated with the oxidative phenotype and capillarity of the muscleand these phenotype-associated modules were enriched for genes involved in energy metabolism muscle plasticityvascular development and cell stress response Although most of the individual transcripts that were differentiallyexpressed between populations were negatively correlated with muscle phenotype several genes involved in energymetabolism (eg Ckmt1 Ehhadh Acaa1a) and angiogenesis (Notch4) were more highly expressed in highlanders and theregulators of mitochondrial biogenesis PGC-1a (Ppargc1a) and mitochondrial transcription factor A (Tfam) were pos-itively correlated with muscle oxidative phenotype These results suggest that evolved population differences in theoxidative capacity and capillarity of skeletal muscle involved expression changes in a small suite of coregulated genes

Key words capillarity hypoxia adaptation muscle fiber type oxygen transport physiological genomics RNA-seq

IntroductionHigh-altitude environments provide fertile ground for exam-ining the integrative physiological and genomic mechanismsof adaptation Patterns of animal diversity change substan-tially across elevational gradients (Sanchez-Cordero 2001McGuire et al 2014) and a high degree of species turnoveroccurs in part from adaptation to local environments(Jankowski et al 2012) The concurrent declines in tempera-ture and oxygen tension with elevation are a particular chal-lenge to small highland mammals which must sustain highrates of aerobic metabolism to support thermogenesis andlocomotion in spite of a diminished oxygen supply (Hayes1989) Both genotypic specialization and phenotypic plasticityin the physiological systems that mediate oxygen transportand utilization could be important for meeting this challengein highland natives but we are just beginning to understandthe mechanisms involved (Storz Scott et al 2010 Scott 2011Cheviron and Brumfield 2012)

Evolved changes in skeletal-muscle phenotype haveoccurred in several highland vertebrate taxa The locomotorymuscle of high-altitude birds that are sampled in their nativeenvironment (Leon-Velarde et al 1993 Hepple et al 1998Mathieu-Costello et al 1998) or raised in captivity at sea level(Scott Egginton et al 2009 Scott Richards et al 2009) istypically highly capillarized and more oxidative than that oflowland birds Highland mammals also exhibit increased ox-idative capacity in the locomotory respiratory andor cardiacmuscles (Sheafor 2003 Cheviron et al 2012 2014 Schipperset al 2012) These derived trait changes should increase thecapacities for oxygen diffusion and aerobic metabolism in themuscle and may therefore enhance fitness-related physiolog-ical performance under hypoxia

Changes in gene regulation have been shown to contributeto evolved differences in locomotory muscle phenotypes thatare related to organismal performance For example miceselected for high levels of voluntary wheel running exhibit

The Author 2015 Published by Oxford University Press on behalf of the Society for Molecular Biology and Evolution All rights reserved For permissions pleasee-mail journalspermissionsoupcom

Mol Biol Evol doi101093molbevmsv076 1

MBE Advance Access published April 17 2015 at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

increased aerobic exercise performance and possess moreoxidative and capillarized gastrocnemius muscle (Audetet al 2011 Templeman et al 2012) which is associatedwith differential expression of many genes involved inmuscle plasticity and calcium signaling (Burniston et al2013) Similarly rats selected for high running enduranceexhibit upregulation of several genes involved in lipid meta-bolism in comparison to low capacity runners (Bye et al2008) In high-altitude populations of deer mice an enhancedthermogenic capacity in hypoxia is associated with higheractivities of oxidative enzymes as well as both constitutiveand plastic changes in gene expression in locomotory musclecompared with lowland deer mice (Cheviron et al 20122014) Although changes in capillarity and fiber compositionof skeletal muscle have evolved in several highland taxa(Leon-Velarde et al 1993 Scott Egginton et al 2009) thetranscriptomic underpinnings of these convergent changesin phenotype are not known

The deer mouse (Peromyscus maniculatus) has thebroadest altitudinal distribution of any North Americanmammal stretching from below sea level in Death ValleyCalifornia to over 4300 m above sea level in numerousmountain ranges (Hock 1964) Population genetic studiesof deer mice in western North America have demon-strated that highland mice are often genetically distinctfrom lowland conspecifics and estimates of gene flowbetween highland and lowland populations are very lowacross the interface between the western Great Plains andthe Front Range of the Southern Rocky Mountains(ltlt001 migrants per generation) (Storz et al 2012Natarajan et al 2015) Survivorship studies of high-alti-tude deer mice have documented strong directionalselection for increased thermogenic capacity under hyp-oxia (Hayes and OrsquoConnor 1999) Such selection is likely tohave influenced the metabolic capacities of skeletalmuscle because shivering thermogenesis requires highlevels of muscle activity (Oufara et al 1987 Van Santand Hammond 2008)

The objective of this study was to determine whether theevolved increases in aerobic capacity and oxidative enzymeactivities in the skeletal muscle of highland deer mice(Cheviron et al 2012 2014) are associated with a more capil-larized and oxidative muscle phenotype We then aimed todetermine the relationship between various phenotypic traitsand gene expression profiles in the skeletal muscle usingwhole transcriptome shotgun sequencing (RNA sequencing[RNA-seq]) in order to elucidate the regulatory changes thatunderlie adaptive modifications of muscle physiology in high-altitude mice Because induced changes in muscle phenotypemay occur at different ontogenetic stages we used an exper-imental design that allowed us to account for the combinedeffects of developmental plasticity and physiological plasticityduring adulthood We made comparisons between highland(Mt Evans CO 4350 m) and lowland (Lincoln NE 430 m)mice that were sampled at their native elevations and wealso compared F1 progeny of wild-caught mice that wereborn and reared under common garden conditions at lowelevation

Results

Muscle Phenotype Differs between Highlanders andLowlanders

High-altitude mice exhibited an enhanced oxidative pheno-type of the locomotory (gastrocnemius) muscle (figs 1 and 2)This result was reflected in significant main effects of popu-lation altitude on several oxidative phenotypes including theareal and numerical densities of oxidative fibers In pairwisecomparisons between populations highlanders exhibitedgreater areal and numerical densities of oxidative fibers(both slow oxidative fiber density as well as total oxidativefiber density) when mice were sampled in their native envi-ronment and these differences were mirrored by nonsignifi-cant trends in the F1 progeny of wild-caught mice that werereared in the common-garden lab environment The totalproportion of oxidative fibers in the muscle was greater inhighland deer mice by approximately 14ndash25 as a proportionof the total transverse area of the muscle and it was greater byapproximately 5ndash14 as a proportion of total fiber number(figs 1 and 2) The main cause of these differences was a 23ndash83 higher abundance by area of slow oxidative (type I)muscle fibers (figs 1 and 2) with no change in the abundanceof fast oxidative (type IIa) muscle fibers (table 1) The popu-lation differences in areal density were generally greater thanthe differences in numerical density because slow oxidativefibers were larger in highland deer mice There were no dif-ferences in the size of fast oxidative and fast glycolytic fibersbetween highland and lowland mice (table 1)

Highland mice also had an increased capillarity in the gas-trocnemius muscle in comparisons involving both native andcaptive-reared individuals Many indices of capillarity werehigher in highland mice compared with lowland mice includ-ing capillary surface density (~23ndash37 higher) capillary den-sity (~10ndash21 higher) the ratio of capillary surface to musclefiber surface (~19ndash36 higher) and the capillary to fiber ratio(~8ndash17 higher) (figs 1 and 3) There was also a noticeablyhigher vessel tortuosity in the highland mice as reflected bythe greater discrepancy in capillary surface density than incapillary density and the qualitative difference in the patternof capillary staining (fig 3) The overall difference in capillaritybetween highland and lowland mice was greater than thedifference between native and F1 mice with the same popu-lation-of-origin indicating that the observed population dif-ference is not attributable to phenotypic plasticity

There was a positive correlation between capillary surfacedensity and the areal density of oxidative fibers (fig 4A) Thusthe increase in capillarity in highland deer mice could haveresulted solely from the greater oxygen demands imposed bythe increase in oxidative fibers in highland mice and not toprovide an additional improvement of oxygen diffusion ca-pacity under hypoxia To explore this possibility we used theresiduals from the regression of capillary surface densityagainst the areal density of oxidative fibers to test for differ-ences in capillarity while controlling for the difference in fibercomposition between highland and lowland deer miceResiduals from this regression were significantly greater in

2

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

highlanders than in lowlanders (fig 4B) suggesting that theenhanced capillarity of highland mice should increase oxygendiffusion capacity in hypoxia

Population Differences and Plasticity in the MuscleTranscriptome

We used a general linear model to identify genes that ex-hibited persistent expression differences between highlandand lowland populations across rearing environments andthose that exhibited significant plasticity (as revealed bywithin-population comparisons between mice sampled intheir native environment vs F1 progeny of wild-caughtmice) In total 657 genes (54 of measured transcripts) ex-hibited significant expression differences between popula-tions (highland vs lowland) andor between rearingenvironments (native vs lab F1) Of these variable transcriptsthe vast majority exhibited significant plasticity in expressionas indicated by differences between rearing environmentswithout a significant population effect (897mdash589 genes)

A smaller number of genes were differentially expressed be-tween populations without significant effects of rearing en-vironment (103mdash68 genes) Twenty-nine genes (44 ofvariable transcripts) exhibited significant expression differ-ences between populations and between rearing environ-ments We focused on this subset of 657 variabletranscripts to investigate the transcriptomic basis of popula-tion differences in muscle phenotypes (supplementary tablesS1 and S2 Supplementary Material online)

We calculated correlation coefficients for all pairwise com-parisons of transcript abundance and we then used modu-lated modularity clustering (MMC) (Stone and Ayroles 2009)to identify transcriptional modules of coregulated genes(Rockman 2008 Ayroles et al 2009) This analysis revealed ahigh degree of correlational structure among differentiallyexpressed transcripts (fig 5) The 68 genes with significantpopulation effects clustered into two modules the largest ofwhich (module P2) contained 50 genes (fig 5A) We recov-ered a similar pattern for the genes with significant differencesin expression between rearing environments (ie native vs lab

FIG 1 Histological analysis of fiber type and capillarity in the gastrocnemius muscle of deer mice Representative images from individuals sampled intheir native environment are shown Oxidative muscle fibers were identified by staining for succinate dehydrogenase activity slow oxidative (type I)muscle fibers were identified by staining for slow myosin ATPase protein using immunohistochemistry and capillaries were identified by staining foralkaline phosphatase activity There were clear differences in staining intensity between highland and lowland deer mice

3

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

F1) (fig 5B) The 589 genes with significant effects of rearingenvironment clustered into a total of five modules the largestof which (module T5) contained nearly 38 of the environ-mentally sensitive transcripts (224 genes)

Relationships between Muscle Phenotype andTranscriptomic Variation

Population-specific differences in gene expression weresignificantly associated with the observed differences in

muscle phenotype We used a combination of principal com-ponents analysis (PCA) and regression analysis to determinewhich transcriptional modules were most strongly associatedwith muscle phenotypes Because genes within transcrip-tional modules are by definition highly correlated PCAcan be used to summarize expression patterns of particulartranscriptional modules (Cheviron et al 2014 Stager et al2015) For each of the transcriptional modules the firstprincipal component (PC1) accounted for 324ndash997of the variance in gene expression among individuals

FIG 2 The gastrocnemius muscle has a more oxidative phenotype in highland deer mice There were significant effects of population altitude on theareal density of oxidative fibers (AA(oxm) area of oxidative fibers relative to the total transverse area of the muscle) (F[137] = 1337 Plt 0001) thenumerical density of oxidative fibers (NN(oxm) number of oxidative fibers relative to the total number of fibers) (F[137] = 5143 P = 0029) and the areal(AA(type Im) F[137] = 9329 P = 0004) and numerical (NN(type Im) F[137] = 6379 P = 0016) densities of slow oxidative fibers The effects of rearingenvironment (native vs common-garden F1 raised in the lab) were not significant (AA(oxm) F[137] = 1168 P = 0287 NN(oxm) F[137] = 1708 P = 0199AA(type Im) F[137] = 0133 P = 0718 NN(type Im) F[137]lt 0001 P = 0990) The interactions between population and rearing environment were alsonot significant (AA(oxm) F[137] = 0694 P = 0410 NN(oxm) F[137] = 1032 P = 0316 AA(type Im) F[137] = 2259 P = 0141 NN(type Im) F[137] = 0724P = 0400) Significant pairwise difference between highlanders and lowlanders within an experimental group (native vs F1) Native lowlanders n = 12F1 lowlanders n = 9 native highlanders n = 9 F1 highlanders n = 11

Table 1 Fiber Types in the Gastrocnemius Muscle of Deer Mice

Variable Tr Lowlanders Highlanders F P

AA(type IIam) Native 0299 0024 0288 0041 Pop 0042 0838Lab F1 0316 0027 0340 0029 RE 1308 0260

NN(type IIam) Native 0353 0030 0328 0044 Pop 0556 0461Lab F1 0383 0023 0359 0032 RE 0885 0353

Type I area Native 1082 114 1503 127 Pop 5092 0030Lab F1 1329 65 1436 136 RE 0593 0446

Type IIa area Native 1448 114 1574 134 Pop 0948 0337Lab F1 1491 115 1598 110 RE 0077 0783

Type IIb area Native 2277 178 2404 130 Pop 0152 0699Lab F1 2561 109 2313 161 RE 0380 0541

NOTEmdashTransverse area of each fiber type (slow oxidative type I fast oxidative type IIa fast glycolytic type IIb) is reported in mm2 AA(IIam) areal density of fast oxidative fibersNN(IIam) numerical density of fast oxidative fibers one degree of freedom for each main effect variable (Pop population altitude RE rearing environment) and 36 for theresidual

Significant pairwise difference between highlanders and lowlanders within an environment

4

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

(supplementary table S3 Supplementary Material online)Therefore we used PC1 scores as an index of overallmodule expression to identify transcriptional modules thatwere significantly associated with muscle phenotypes Of thetwo modules composed of genes with significant populationeffects only one (P2) was significantly associated with any ofthe measured phenotypic traits Module P2 expression scoreswere negatively associated with the total areal (R2 = 051P = 000026) and numerical (R2 = 029 P = 00081) densitiesof oxidative fibers capillary density (R2 = 023 P = 00223) cap-illary surface density (R2 = 046 P = 00009) and the ratio ofcapillary surface area to fiber surface area (R2 = 051P = 00004) (table 2) The relationship between module P2expression scores and capillary to fiber ratio were marginallysignificant (R2 = 016 P = 00513) Correspondingly highlandmice exhibited significantly lower module P2 expressionscores than lowland mice regardless of rearing environment(fig 6) PC2 scores for module P1 were associated with severalphenotypic traits but this axis accounted for a trivial amountof the gene expression variance within this module (02)(supplementary table S3 Supplementary Material online)

Of the five environmentally sensitive transcriptional mod-ules (ie those composed of genes exhibiting a significanteffect of rearing environment) only one was statistically as-sociated with several phenotypic traits Module T5 expression

was positively associated with the areal (R2 = 030 P = 0007)and numerical (R2 = 032 P = 0005) densities of slow oxidativefibers (table 2) and expression differences were highly signif-icant in the comparison between highland and lowland micein their native environments (fig 6) PC2 scores for modulesT1 and T2 were associated with several phenotypic traits butagain this axis accounted for little of the explained variance ingene expression within modules (33 and 56 respectively)(supplementary table S3 Supplementary Material online)

Enrichment analyses revealed that the transcriptionalmodules associated with muscle phenotypes contain a di-verse assemblage of genes with a variety of molecular func-tions many of which are relevant for metabolism musclefiber-type differentiation and angiogenesis (supplementarytable S4 Supplementary Material online) Module P2mdashthelarger module comprising genes that were differentially ex-pressed between populations regardless of rearing environ-mentmdashwas enriched (at a P value corrected for the falsediscovery rate [FDR] qlt 010) for genes annotated withfour gene ontology (GO) terms associated with metabolicprocesses ldquosecondary metabolic processrdquo (q = 0034) ldquooxida-tionndashreduction processrdquo (q = 0064) ldquotoxin metabolic pro-cessrdquo (q = 0016) and ldquoorganic acid metabolic processrdquo(q = 0094) Module T5 was significantly enriched for genesannotated with several GO terms related to muscle

FIG 3 The gastrocnemius muscle has a higher capillarity in highland deer mice than in lowland deer mice There were significant effects of populationaltitude on capillary surface density (CSD mm of capillary surface per mm2 of transverse muscle area) (F[135] = 1892 Plt 0001) the ratio of capillarysurface to fiber surface (CSFS) (F[135] = 1920 Plt 0001) the density of capillaries (CD capillaries per mm2 of transverse muscle area) (F[135] = 4525P = 0041) and the number of capillaries per muscle fiber (CF) (F[135] = 4223 P = 00474) The effects of rearing environment (native vs common-gardenF1 raised in the lab) were not significant (CSD F[135] = 0781 P = 0383 CSFS F[135] = 0071 P = 0792 CD F[135] = 1688 P = 0202 CF F[135] = 0517P = 0477) The interactions between population and rearing environment were also not significant (CSD F[135] = 0942 P = 0339 CSFS F[135] = 1337P = 0255 CD F[135] = 0321 P = 0575 CF F[135] = 0430 P = 0516) Significant pairwise difference between highlanders and lowlanders within anexperimental group (native vs F1) Native lowlanders n = 12 F1 lowlanders n = 9 native highlanders n = 8 F1 highlanders n = 10

5

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

phenotype and oxygen transport including ldquoblood vessel de-velopmentrdquo (q = 0016) ldquomuscle structure developmentrdquo(q = 0020) ldquovascular developmentrdquo (q = 0028) and ldquonegativeregulation of vascular permeabilityrdquo (q = 0045) Module T5was also enriched for many genes involved in energy metab-olism cell signaling cell migration and cell turnover ModulesT2 T3 and T4 were significantly enriched for genes thatparticipate in immune function apoptosis and cell stress re-sponse but these modules were not significantly associatedwith any of the measured muscle phenotypes

Expression levels of 27 individual genes within modules P1and P2 were greater in highlanders than in lowlanders

corresponding with the observed differences in muscle phe-notype (fig 7 and supplementary table S1 SupplementaryMaterial online) Highland deer mice exhibited higher expres-sion of several genes that are associated with oxidative energymetabolism including enoyl-coA hydratase (Ehhadh) andacetyl-coenzyme A acyltransferase (Acaa1a) (enzymes in-volved in -oxidation) mitochondrial creatine kinase(Ckmt1) a mitochondrial ribosomal protein (Mrpl22) andmannose receptor C type 1 (Mrc1) Expression of the latteris positively associated with mitochondrial gene expression inthe muscle of humans (Moreno-Navarrete et al 2013)Highlanders also exhibited higher expression of two aldehydedehydrogenase paralogs (Aldh1a1 and Aldh1a7) whichencode enzymes that can protect against ischemia-reperfu-sion injury by reducing oxidative stress autophagy and apo-ptosis during hypoxia (Ma et al 2011 Contractor et al 2013Zhang et al 2013) The expression of two genes that havebeen associated with angiogenesis cadherin-7 (Cdh7)(Hayward et al 2011) and Notch-4 (Notch4) (Lv et al 2013)was also more highly expressed in highlanders Most of theremaining genes with high expression in highlanders areuncharacterized (supplementary table S1 SupplementaryMaterial online)

Most of the genes that were differentially expressedbetween populations were expressed at a lower level in thehighland mice (supplementary table S1 SupplementaryMaterial online) Individual genes that encode proteins in-volved in the sarcomere (Actn1) peptide transport(Slc15a2) and muscle growth (Trf) were expressed at lowerlevels in highland mice possibly because expression of thesegenes is more prevalent in fast glycolytic fibers (which are lessabundant in the gastrocnemius of highlanders figs 1 and 2)Several transcripts for genes involved in cellular protectionfrom reactive O2 species or xenobiotics were also lower inhighlanders such as transferrin (Trf) ceruloplasmin (CP) glu-tathione S-transferase (Gstt3) cytochrome P450 enzymes(9030605E09Rik) and flavin-containing monooxygenase(Fmo1) Curiously some genes involved in oxidative energymetabolism (Atp6 ND3 ND4 Ndr1 and Pank1) as well as apotent vasodilator of arteries (urocortin Ucn) (Diaz andSmani 2013) were also expressed at lower levels inhighlanders

Relationships between Muscle Phenotype and theExpression of Candidate Genes

In addition to the discovery-driven analyses of transcriptionalvariation we tested for associations between muscle pheno-type and expression level for a number of candidate genesinvolved in regulating energy metabolism mitochondrial bio-genesis andor angiogenesis (Lin et al 2005 Lanza andSreekumaran Nair 2010 Gustafsson 2011) (see Materialsand Methods) There were functional similarities betweenour set of candidate genes and the set of genes that wereoverrepresented in the enrichment analyses which is notsurprising given that both sets of genes are associated withmetabolism muscle fiber-type differentiation and angiogen-esis Six of the candidate genes exhibited positive associations

FIG 4 Capillarity in the gastrocnemius muscle is greater in highlanddeer mice than expected from the variation in muscle oxidative phe-notype (A) There was a strong linear correlation between capillarysurface density (CSD) and the areal density of oxidative fibers(AA(oxm)) (CSD = 00540 AA(oxm) + 00133 Plt 0001) Dashed linesrepresent the 95 confidence intervals of the regression Symbols areas follows F1 lab-raised lowlanders black upwards triangles native low-landers white upwards triangles F1 lab-raised highlanders dark graydownwards triangles native highlanders light gray downwards triangles(B) There was a significant effect of population altitude on the residualCSD from the regression in (A) (F[135] = 5558 P = 0024) but there wasno significant effect of rearing environment (F[135] = 2453 P = 0126)and no significant interaction between population and rearing environ-ment (F[135] = 0337 P = 0566)

6

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

with at least one phenotypic trait Expression of the peroxi-some proliferator-activated receptors (PPAR) g coactivator1a (PGC-1a) gene Ppargc1a a purported master regulatorof mitochondrial biogenesis was associated with the numer-ical density of oxidative fibers in the muscle (R2 = 0203P = 0046) and the ratio of capillary surface to fiber surface(R2 = 0239 P = 0034) (table 3) The association betweenTfam a key activator of mitochondrial gene expression andthe areal density of oxidative fibers was marginally significant

(R2 = 0197 P = 0050) Several of the growth factors (Angpt1Pdgfa Pdgfd and Vegfc) were positively associated withthe size of a particular fiber type (table 3) (althoughonly slow oxidative fiber area differed between populationstable 1)

The majority of significant associations between candidategene expression and muscle phenotype was negative(table 3) Fifteen of the 43 candidate genes showed significantnegative associations with at least one phenotypic trait

FIG 5 Correlated transcriptional modules (A) Clustering of the 68 transcripts with significant population effects into two transcriptional modules(B) Clustering of 589 transcripts with significant effects of rearing environment into five transcriptional modules

7

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

(Fgf12 Fgf13 Fgf14 Fgfr2 Mmp2 Mmp13 Mmp14 PdgfcPdgfra Ppara Pparg Tie1 Vegfa Vegfb and Vegfc) The stron-gest negative association was between Fgfr2 and the numer-ical (R2 = 0574 P = 00001) and areal (R2 = 0446 P = 00013)densities of slow oxidative fibers in the muscle (fig 8)

DiscussionDeer mice at high altitudes sustain high metabolic rates tosupport locomotion and thermogenesis (Hayes 1989) andenhanced performance capacities appear to be adaptive(Hayes and OrsquoConnor 1999) Highland deer mice have there-fore evolved an elevated aerobic capacity in hypoxia relativeto their lowland counterparts (Cheviron et al 2012 2014)Here we show that this is partly due to a substantial increasein capillarity and oxidative fiber abundance in the skeletalmuscle We also show that the derived muscle phenotypeof highland deer mice is associated with significant variationin gene expression across the skeletal muscle transcriptomeFor several candidate genes population differences in expres-sion levels were associated with specific muscle phenotypesand many of the differentially expressed genes are known toplay a role in energy metabolism muscle fiber compositionand vascular development

High-Altitude Adaptation and Muscle Phenotype

The highly oxidative phenotype of skeletal muscle in highlanddeer mice (figs 1 and 2) which occurs in conjunction with anincrease in the activity of several oxidative enzymes (Chevironet al 2012 2014) could have multiple potential benefits athigh altitudes Highland animals must cope with cold tem-peratures so the evolution of a more oxidative muscle phe-notype could enhance the capacity for shivering and possiblynonshivering thermogenesis (Mineo et al 2012) A highly ox-idative muscle phenotype should also increase the total mi-tochondrial respiration of an entire muscle when intracellularO2 tensions fall (Hochachka 1985 Scott Richards et al 2009)This is because mitochondrial respiration can be limited by O2

during intracellular hypoxia (Gnaiger 2001) which wouldreduce the maximum attainable respiration of individualmuscle fibers Therefore having more oxidative fibersshould counterbalance the inhibitory effects of hypoxia onindividual muscle fibers The adaptive significance of having amore oxidative phenotype could involve one or both of thesemechanisms in highland taxa that have evolved this derivedtrait (Leon-Velarde et al 1993 Mathieu-Costello et al 1998Scott Egginton et al 2009)

The augmented capillarity in the locomotory muscle ofhighland mice (figs 1 and 3) should increase the diffusioncapacity for oxygen from the blood and should thereforeconfer an advantage for sustaining aerobic performance inhypoxia (Wagner 1996 Scott and Milsom 2006 Cano et al2013) There is generally a strong relationship between themitochondrial oxygen demands of a tissue and the capacityfor oxygen supply from the microcirculation (Hepple 2000)Interestingly the magnitude of the increase in capillarity inhighlanders appears to be greater than the increase in oxida-tive capacity (fig 4) This suggests that oxygen diffusion ca-pacity is further enhanced in highlanders to improve oxygentransport in hypoxia This should also be advantageous inhighland deer mice given that they have evolved an increasedhemoglobin-O2 affinity (Chappell and Snyder 1984 Storzet al 2009 Storz Runck et al 2010 Natarajan et al 20132015) as the increased muscle diffusion capacity mitigates the

FIG 6 Altitudinal variation in the expression of transcriptional modulesthat are statistically associated with muscle phenotypic traits Moduleexpression was summarized using PCA and PC1 scores are shown Therewere significant effects of population altitude on modules P2(F[117] = 2380 Plt 0001) and T5 (F[117] = 1019 P = 0005) There wasalso a significant effect of rearing environment (native vs common-garden F1 raised in the lab) on module T5 (F[117] = 2964 Plt 0001)but not module P2 (F[117] = 0937 P = 0347) Transcriptional modulesare shown in figure 5 and the genes that compose each module arepresented in supplementary tables S1 and S2 Supplementary Materialonline Significant pairwise difference between highlanders and low-landers within an experimental group (native vs F1) Native lowlandersn = 6 F1 lowlanders n = 5 native highlanders n = 5 F1 highlanders n = 5

Table 2 Transcriptomic Modules Associated with MusclePhenotypes

Trait Associated modules

Total areal density of oxidative fibers P2 T5

Total numerical density of oxidative fibers P2 T5

Areal density of fast oxidative fibers T5

Numerical density of fast oxidative fibers T5

Capillary density P2

Capillary surface density P2

Capillary surface per fiber surface P2

NOTEmdashThe genes that comprise each module are presented in supplementary tablesS1 and S2 Supplementary Material online

Module phenotype associations that remain significant after Bonferroni correctionfor multiple tests (Plt 000045)

8

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

trade-off between pulmonary O2 loading (which is importantfor safeguarding arterial O2 saturation under hypoxia) and O2

unloading in the peripheral circulationDespite considerable differences in a range of environmen-

tal factors between the two rearing environments (eg tem-perature ambient partial pressure of O2 humidity food typefood availability and biotic interactions) population differ-ences in capillarity persisted in comparisons of both nativeand lab-reared F1 individuals Similarly the difference in theproportional abundance of oxidative fibers was generallymore pronounced in comparisons between highland andlowland deer mice than in comparisons between nativeversus lab-reared F1 individuals with the same population-of-origin although this trait may have also been affected byrearing environment (as revealed by a lack of significant pair-wise differences between populations in the F1 mice)Although we cannot entirely rule out transgenerational epi-genetic effects these results suggest that genetic differencesbetween highland and lowland deer mice may underlie adap-tive variation in muscle phenotype It remains possible that

these traits can be further modified by phenotypic plasticityhowever the effects of cold acclimation on oxidative capacityare not consistent across studies of domestic mice Increasesin the activity of oxidative enzymes in the muscle occur inresponse to cold acclimation in some studies (Mineo et al2012) but not others (Beaudry and McClelland 2010) and therespiratory capacity of isolated mitochondria does not seemto be altered by cold acclimation (Meyer et al 2010 Mineoet al 2012) Likewise combining cold and hypoxia has noeffect on the activity of oxidative enzymes in the gastrocne-mius muscle in domestic mice (Beaudry and McClelland2010) Furthermore muscle capillarity and oxidative capacityoften do not change or even decrease in response to high-altitude hypoxia exposure in lowland humans and othermammals (Mathieu-Costello 2001 Levett et al 2012 Jacobset al 2013) Although muscle disuse (eg limb immobiliza-tion) is associated with substantial changes in muscle pheno-type (Clark 2009) some wild rodents experience minimalchanges in muscle form and function in response to largeseasonal variations in activity during hibernation or torpor

FIG 7 Some individual genes involved in metabolism and angiogenesis were more highly expressed in highlanders than in lowlanders Reaction normsfor gene expression are shown with native environment on the left and lab environment on the right and data are shown relative to the averagenormalized read count for native lowlanders There was a statistically significant effect of population altitude on all genes shown (see supplementarytable S1 Supplementary Material online) The mean normalized read counts (cpm) for native lowlanders were as follows enoyl-CoA hydratase(Ehhadh) 37 acetyl-Coenzyme A acyltransferase 1A (Acaa1a) 45 mitochondrial creatine kinase 1 (Ckmt1) 29 mitochondrial ribosomal proteinL22 (Mrpl22) 85 aldehyde dehydrogenase 1A1 (Aldh1a1) 418 aldehyde dehydrogenase 1A7 (Aldh1a7) 202 mannose receptor C type 1 (Mrc1) 237cadherin-7 (Cdh7) 38 and Notch-4 (Notch4) 458 There were also significant effects of rearing environment on Aldh1a1 and Cdh7 (y) (see supple-mentary table S2 Supplementary Material online) n = 6 for all groups

Table 3 Candidate Genes that Were Associated with Muscle Phenotypes

Trait Positive Association Negative Association

Areal density of oxidative fibers Tfam Fgf12 Pdgfra Vegfc

Numerical density of oxidative fibers Ppargc1a Fgf12 Mmp2 Pdgfra Pparg

Areal density of slow oxidative fibers Fgfr2 Mmp2 Mmp14 Pdgfra Tie1

Numerical density of slow oxidative fibers Fgfr2 Mmp2 Mmp13 Mmp14 Pdgfra Tie1

Type I fiber area Angpt1

Type IIa fiber area Pdgfa Fgf13 Fgf14

Type IIb fiber area Pdgfd Vegfc Fgf14

Capillary density Pdgfc Ppara Vegfa

Capillary to fiber ratio Fgfr2 Mmp2 Mmp14 Pdgfra

Capillary surface density (CSD) Fgfr2 Ppara Vegfb

Capillary surface per fiber surface Ppargc1a Fgfr2

Residual CSD Ppara Vegfa Vegfb

NOTEmdashResidual CSDs are those calculated in figure 4

Association that remained significant after Bonferroni correction for multiple comparisons (Plt 00001)

9

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

(Cotton and Harlow 2010) Therefore evolved genotypic spe-cializations may play a more important role than phenotypicplasticity in creating the muscle phenotype of nativehighlanders

Transcriptomic Basis of the High-Altitude Phenotype

The evolved differences in gene expression between highlandand lowland populations suggest that regulatory changes ingenes involved in aerobic metabolism have contributed tohigh-altitude adaptation (fig 7) Expression of the transcrip-tional coactivator PGC-1a (Ppargc1a) which activates the

expression of genes involved in mitochondrial metabolism(Lin et al 2005 Lanza and Sreekumaran Nair 2010) was pos-itively associated with the abundance of oxidative fibers in themuscle (table 3) and was expressed at higher levels in high-landers than lowlanders Mitochondrial creatine kinase(Ckmt1) which is prevalent in slow oxidative fibers and isimportant for controlling oxidative phosphorylation to pro-mote energy supply-demand coupling and metabolite stabil-ity (Hochachka 1993 Ventura-Clapier et al 1998) wasexpressed at higher levels in highlanders than lowlandersThe potential importance of this enzyme in regulating mito-chondrial respiration under hypoxia is supported by experi-mental studies in humans (Ponsot et al 2006) rats (Walshet al 2006) and birds (Scott Richards et al 2009) The mito-chondrial isoform of creatine kinase is associated with mito-chondria and is highly expressed in tissues with high ATPdemands (Wallimann et al 1992) so its upregulation in high-landers could be associated with the observed variation inexpression of Ppargc1a and Tfam Genes that encode themetabolic enzymes acetyl-coenzyme A acyltransferase(Acaa1a) and aldehyde dehydrogenase 1A1 (Aldh1a1)which are upregulated in mice selected for high levels of vol-untary wheel running (Burniston et al 2013) were also morehighly expressed in highlanders than in lowlanders Expressionof Acaa1 is known to be induced by PPAR transcription fac-tors (Guo et al 2007) for which PGC-1a is a coactivatorsuggesting a potential relationship between the patterns ofvariation in expression of Ppargc1a and Acaa1a However notall genes involved in oxidative energy metabolism were upre-gulated Some mitochondrially encoded transcripts includingtwo of complex I of the electron transport system (ND3 andND4 transcripts of two of the 44 subunits of NADH dehy-drogenase) and one of ATP synthase (Atp6 one of nine sub-units of the FO region of ATP synthase) were downregulatedin highlanders The meaning of these changes in expression isunclear because NADH dehydrogenase and ATP synthase aremultisubunit enzymes whose functions require coordinatedexpression of all polypeptide subunits (Duggan et al 2011Suarez and Moyes 2012) Nevertheless our findings suggestthat high-altitude adaptation involves substantial transcrip-tomic restructuring in association with a more oxidativemuscle phenotype

Changes in the expression of genes involved in regulatingthe capillary network also appear to have contributed toevolved phenotypic differences between highland and low-land populations Cadherin-7 and Notch-4 in particular areexpressed at higher levels in highlanders than in lowlandersAlthough cadherin-7 is a N-type cadherin that is found pri-marily in neural tissue (Faulkner-Jones et al 1999) its expres-sion increases in association with angiogenesis in response toischemia in the brain (Hayward et al 2011) and E-type cad-herins are known to be extremely important cell adhesionmolecules in angiogenesis The Notch signaling pathway isimportant for coordinating cell behaviors that lead to thecreation of stable and patent blood vessels during angiogen-esis and the notch-4 receptor is restricted to the vascularsystem (Sainson and Harris 2008 Lv et al 2013) Whennotch-4 is bound by its ligand (delta-like 4) signaling by the

FIG 8 There was a strong negative association between the expressionof fibroblast growth factor receptor 2 (Fgfr2) and the abundance of slowoxidative fibers There was a significant linear correlation between Fgfr2transcript abundance and both the numerical (R2 = 0574 P = 00001)and areal (R2 = 0446 P = 00013) densities of slow oxidative fibers in thegastrocnemius muscle Dashed lines represent the 95 confidence in-tervals of the regression Symbols are as follows F1 lab-raised lowlandersblack upwards triangles native lowlanders white upwards triangles F1lab-raised highlanders dark gray downwards triangles native high-landers light gray downwards triangles

10

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

receptor inhibits vascular endothelial growth factors (VEGF)expression as well as endothelial cell proliferation and migra-tion (Williams et al 2005) This is notable in light of the neg-ative association we observed between capillarity and theexpression of Vegfa and other growth factors involved in pro-moting the proliferative and migratory stages of angiogenesis(table 3) It is possible that the highly capillarized phenotypeof skeletal muscle in highlanders is initiated by growth factorsignaling at an earlier stage of development and that it ismaintained during adulthood by an upregulation of notch-4 expression to stabilize the mature capillary network

There was a negative association between muscle pheno-types and the expression of many genes PC1 scores ofmodule P2 expression were lower in highlanders than in low-landers (fig 6) (corresponding to lower abundance in high-landers for roughly half of the transcripts in module P2) andwere negatively associated with several of the oxidative andcapillarity phenotypes A similar observation was made forPC1 of module T5 expression Collectively these modules areenriched for genes assigned to several GO processes involvedin metabolism muscle plasticity and vascular developmentThe majority of significant associations between muscle phe-notypes and the expression of candidate genes involved inenergy metabolism mitochondrial biogenesis or angiogenesiswas also negative (table 3) These results were surprising andunexpected because the protein products of many of thegenes that were expressed at lower levels in highlandershave been shown to contribute positively to capillarity oroxidative capacity in rodents and other mammals (Lin et al2005 Lanza and Sreekumaran Nair 2010 Gustafsson 2011)These seemingly paradoxical results may result from the ef-fects that differences in muscle phenotype should have onintracellular signals of oxygen and energy homeostasis Forexample differential expression of a small number of othergenes (such as Notch-4) could maintain the highly capillarizedphenotype of highlanders which would increase cellular O2

supply relative to lowlanders and would presumably dampensignaling by the factors that drive the responses to oxygenlimitation (such as hypoxia-inducible factors and VEGF) Anegative association between muscle phenotypes and expres-sion could thus result for genes that do not cause the differ-ences in muscle phenotype but are sensitive to its effects

As the increased oxidative capacity and capillarity of skel-etal muscle in highland mice were not primarily attributableto phenotypic plasticity we reasoned that the underlyinggenes would not exhibit plasticity in expressionNevertheless a large number of genes were differentially ex-pressed between mice in the native and lab environments(figs 5 and 6) Most of the expression differences were notassociated with any muscle phenotypes and may insteadreflect transcriptomic changes associated with different as-pects of natural and lab environments (table 2) Indeed en-vironmentally sensitive modules that were not associatedwith muscle phenotype were enriched for genes involved insignaling cell migration cell turnover immune function andcell stress which may reflect the substantial differences be-tween native and lab environments in features such as foodaccess diet quality exposure to pathogens and a range of

other factors in addition to differences in O2 tensionDetermining which of these environmentally sensitive genesare responding to changes in elevation will require controlledreciprocal acclimation experiments

ConclusionsOne of the goals of evolutionary physiology is to elucidate themechanistic basis of adaptive variation in organismal perfor-mance (Garland and Carter 1994 Dalziel et al 2009)Populations and species that have diverged across altitudinalgradients are well suited to this endeavor They provide anopportunity to elucidate how genotypic specialization phe-notypic plasticity and their interaction can contribute toadaptive enhancements of physiological performance in chal-lenging environments particularly as cold and hypoxia havewell-defined impacts on the physiological systems importantfor exercise and thermogenesis In this study we show thathighland deer mice have a highly oxidative and capillarizedmuscle phenotype that is associated with population differ-ences in expression for a small suite of genes involved inmetabolism muscle plasticity and vascular developmentThe capacity for oxygen transport and utilization in themuscle is extremely important for thermogenesis (Oufaraet al 1987 Van Sant and Hammond 2008)mdashan organismalperformance trait with demonstrated fitness benefits in high-altitude deer mice (Hayes and OrsquoConnor 1999)mdashsuggestingthat the unique muscle phenotype and transcriptional profileof highland deer mice are adaptive Consistent with previouswork in deer mice (Cheviron et al 2012 2014) these resultssuggest that differences in both evolved and plastic gene ex-pression patterns contribute to fitness-related variation inphysiological performance at high altitudes by altering phe-notypic traits at several levels of biological organization

Materials and Methods

Populations of Highland and Lowland Deer Mice

Adult deer mice were live trapped on the summit of MountEvans (Clear Creek County CO at 39 3501800N 105 3803800W4350 m above sea level) (P m rufinus) and at low altitude inthe Great Plains (Nine Mile Prairie Lancaster County NE at40 5201200N 96 48020300W 430 m above sea level) (P mnebracensis) as previously described (Cheviron et al 20132014) One set of mice the in situ group from each localitywas sampled on the day of capture at their native elevationfor histology and transcriptomics Another set of mice wastransported to a common-garden lab environment at theUniversity of Nebraska (elevation 360m) and used as a paren-tal stock to establish captive bred highland and lowland linesFrom each of these lines we sampled full-sibling F1 progenyfrom one highland and one lowland breeding pair All F1 micewere born and raised to adulthood in a common normoxicenvironment and were then sampled for histology and tran-scriptomics Lab-raised mice were held in standard holdingconditions at 25 C with unlimited access to mouse chow andwater (12 h light12 h dark photoperiod) All experimentalprotocols were approved by the University of Nebraska

11

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

Institutional Animal Care and Use Committee (IACUCno 522)

Muscle Histology

Muscle capillarity and oxidative phenotype were assessed aspreviously described (Scott Egginton et al 2009 Scott andJohnston 2012) An entire gastrocnemius muscle was dis-sected coated in embedding medium frozen in liquid N2-cooled isopentane and stored at 80 C The muscle wassectioned (10mm) transverse to muscle fiber length in a cryo-stat at 20 C Alkaline phosphatase activity was used toidentify capillaries by staining in assay buffer (concentrationsin mM 10 nitroblue tetrazolium 05 5-bromo-4-chloro-3-in-doxyl phosphate 28 NaBO2 7 MgSO4 pH 93) for 1 h at roomtemperature Succinate dehydrogenase activity was used toidentify oxidative muscle fibers (both slow and fast) also bystaining for 1 h at room temperature (assay buffer concentra-tions in mM 06 nitroblue tetrazolium 20 KH2PO4 154Na2HPO4 167 sodium succinate) Slow myosin immunore-activity using the S58 antibody was used to identify oxidativemuscle fibers (Developmental Studies Hybridoma Bank IowaCity IA) as follows After an initial fixation in acetone for10 min sections were blocked in 10 normal goat serum(made up in phosphate-buffered saline [PBS] [015 mol l1pH 74] containing 1 Triton X-100 and 15 BSA [PBSTXBSA]) for 1 h They were then incubated overnight at 2 C inS58 antibody solution (110 dilution in PBSTXBSA) Thenext morning sections were treated for 10 min withPeroxidase Blocking Reagent (Dako Burlington ONCanada) incubated in secondary antibody (antimousebiotin IgA Southern Biotech Birmingham AL) solution(120 dilution in PBSTXBSA) for 1 h incubated inExtrAvidin-Peroxidase solution (150 dilution in PBSTXBSA Sigma-Aldrich Oakville ON Canada) for 30 min andfinally developed in 04 mg ml1 3-amino-9-ethyl-carbazolesolution (containing 002 H2O2 in 005 M sodium acetatebuffer pH 50) for approximately 5 min Sections were wellrinsed in PBS between each of the above steps Images werecollected using light microscopy and stereological quantifica-tion methods were used to make unbiased measurements(Weibel 1979 Egginton 1990) We analyzed a sufficientnumber of images for each sample to account for heteroge-neity determined by the number required to yield a stablemean value

There is generally a strong relationship between the capac-ity for oxygen diffusion from capillaries and the mitochondrialoxygen demands they are meant to support (Hepple 2000)Therefore our expectation was that differences in muscleoxidative capacity would be associated with correspondingdifferences in muscle capillarity If increases in muscle capil-larity arose in highland mice to enhance O2 transport in hyp-oxia then capillarity would be greater in highland mice thanpredicted by the normal relationship between capillarity andoxidative capacity To assess whether there was variation incapillarity that was independent of the variation in muscleoxidative phenotype we first regressed capillary surface den-sity to the areal density of oxidative fibers and we then used

residuals from this regression for all subsequent statisticalcomparisons (see below)

Histology data are generally reported as means standarderror (except when data points from individual samples areshown) Two-factor analysis of variance and Bonferroni mul-tiple comparisons tests were used as appropriate to assess themain effects and interactions of population and acclimationenvironment and significance level of Plt 005 was used

Muscle Transcriptomics

We utilized previously published gastrocnemius RNA-seqdata (Cheviron et al 2014) to quantify genome-wide patternsof gene expression for 24 individuals that were also used formuscle phenotyping to determine the transcriptomic basis ofvariation in muscle phenotypes (native elevation groupHighland mice n = 6 lowland mice n = 6 Lab F1 groupHighland mice n = 6 lowland mice n = 6) (NCBI SRA acces-sion number SRA091630) These data were originally used toidentify the genomic basis of thermogenic performance dif-ferences between high and low elevation populations In thisstudy we present a reanalysis of a subset of these data to testfor associations between transcriptomic profiles and musclephenotypes (see supplementary table S5 SupplementaryMaterial online for the individuals that were used from thislarger data set) We isolated mRNA from gastrocnemismuscle using a micro PolyA purist kit (Ambion) and gener-ated Illumina sequencing libraries following standard proto-cols (available upon request) Libraries were sequenced as 76nt single-end reads on the Illumina Genome Analyzer IIx fiveindividuals were multiplexed using Illumina index primersand were sequenced in a single lane of a flow cell Imageanalysis and base calling were performed using Illumina pipe-line software Our sequencing strategy produced an averageof 59 million reads per individual (range = 105ndash157 millionreadsindividual)

Read Mapping Normalization and Statistical Analysisof RNA-seq Data

We performed a series of sequential filtering steps to removelow-quality reads and base calls as well as technical artifactsstemming from library preparation First reads with meanPhred quality scores less than 30 were removed from thedata set Second low-quality bases (Phred scorelt 30) weretrimmed from these remaining high-quality sequences usingthe ldquoTrim Sequencesrdquo tool in CLC Genomics Workbench 604(Trimming settings Trim using quality scores limit 0001)Finally reads were scanned for exact matches to the knownIllumina adaptor sequences using the ldquoTrim Adaptorsrdquo tool inCLC Genomics Workbench and if detected they weretrimmed from the sequence read We estimated transcriptabundance by mapping sequence reads to the Mus musculusgenome build 361 using CLC Genomics workbench (map-ping settings minimum length fraction = 09 minimum sim-ilarity fraction = 08 and maximum number of hits for aread = 10) We chose to use the well-annotated and estab-lished Mus genome as reference instead of the draft P man-iculatus genome to avoid issues associated with poor

12

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

assembly and annotations that are common in draft genomeassemblies (Denton et al 2014) but this may have preventedus from estimating expression of rapidly evolving genes thatare highly divergent between Mus and Peromyscus Finally weexcluded genes with less than an average of five reads perindividual because genes with low count values are typicallysubject to increased measurement error and reduced powerto detect differential expression (Robinson and Smyth 2007)Filtering low coverage transcripts has the benefit of increasingstatistical power to detect differential expression for well-sam-pled (ie high coverage) genes by reducing the number ofindependent tests that are performed This filtering strategycould potentially result in bias against short lowly transcribedgenes but we are unlikely have sufficient power to detectdifferential expression of such genes anyway because of theirlow sequence coverage These trimming and filtering stepsresulted in a final data set of 12175 detected genes and anaverage of 26 million sequence reads per individual (range of05ndash67 million reads)

We used the function calcNormFactors in the programedgeR (Robinson et al 2010 Robinson and Oshlack 2010)to normalize read counts among individuals and to controlfor differences in the total library size (number of total reads)among individuals Following this normalization procedurewe tested for differences in transcript abundance betweenpopulations (highland vs lowland) and rearing environments(in situ and F1) using a generalized linear model approach inedgeR Population and rearing environment were included asthe main effects and we included a term for their interactionWe estimated model dispersion for each gene separatelyusing the function estimateTagwiseDisp in edgeR(McCarthy et al 2012) and tested for genes that exhibitedsignificant expression differences between populations andor rearing environments using a general linear model (GLM)likelihood ratio test implemented in edgeR We controlled formultiple tests by enforcing a genome-wide FDR of 005(Benjamini and Hochberg 1995)

We assessed the degree of correlation in transcript abun-dance among the genes with significant population or envi-ronment effects (FDRlt 005 n = 657 genes see Results) todefine putative transcriptional modules of coexpressed genes(Ayroles et al 2009) To do this we calculated Pearson cor-relation coefficients for all pairwise gene expression valuesand then used MMC to group genes according to their pair-wise correlation coefficients (Ayroles et al 2009 Stone andAyroles 2009) To determine which of these transcriptionalmodules were most strongly associated with muscle pheno-types we used PCA to summarize overall module expressionBecause the first two principal component axes (PC1 andPC2) accounted for 47ndash100 of the explained gene expres-sion variance within modules PC1 and PC2 scores were usedas a proxy measure to summarize overall module expressionModule-specific PC1 and PC2 scores were then used in linearregression analyses to test for associations between moduleexpression and each of the measured phenotypes PCA andlinear regressions were performed in R (ver 311R Development Core Team 2014) and we corrected formultiple tests using Bonferroni adjusted P values

Once trait-associated modules were identified we used asuite of functional annotation tools to identify specific ldquobio-logical processrdquo GO terms associated with each module Firstwe used MGI GO Term Mapper (httpwwwinformaticsjaxorgtoolsshtml last accessed March 25 2015) to identify GOSlim terms associated with each gene in a given module thenwe used the MGI GO Term Mapper and GOrilla (Eden et al2009) to test for functional enrichment of specific termswithin modules and we visualized the GOrilla output usingthe program REViGO (Supek et al 2011)

In addition to these discovery-driven enrichment analyseswe tested for associations between muscle phenotype andexpression level for a number of candidate genes (using linearregression analyses as described above) We considered genefamilies involved in regulating energy metabolism andor mi-tochondrial biogenesis including the PPAR (Ppara Ppardand Pparg) the PPARg coactivators (PGC Ppargc1aPpargc1b and Pprc1) nuclear respiratory factors 1 (Nrf1)and 2 (Nfe2l2) and mitochondrial transcription factor A(Tfam) (Lin et al 2005 Lanza and Sreekumaran Nair 2010)We also considered a number of gene families in which one ormore members are involved in angiogenesis including VEGF(Vegfa Vegfb and Vegfc) and their receptors (VEGFR1 Flt1and VEGFR2 Kdr) angiopoietins (Angpt1 Angptl2 andAngptl4) and their receptors (Tie1 and Tek) fibroblastgrowth factors (FGF Fgf1 Fgf6 Fgf7 Fgf10 Fgf12 Fgf13 andFgf14) and their receptors (FGFR Fgfr1 Fgfr2 Fgfr3 and Fgfrl1)platelet-derived growth factors (PDGF Pdgfa Pdgfb Pdgfcand Pdgfd) and their receptors (PDGFR Pdgfra Pdgfrb andPdgfrl) and matrix metalloproteinases (MMP Mmp2 Mmp9Mmp13 Mmp14 Mmp15 and Mmp24) (Gustafsson 2011)We targeted all members of each particular family that wereexpressed in the muscle including genes whose influence onmuscle phenotype has not been previously documented

Supplementary MaterialSupplementary tables S1ndashS5 are available at Molecular Biologyand Evolution online (httpwwwmbeoxfordjournalsorg)

Acknowledgments

The authors thank Colin Nurse Cathy Vollmer and DanielleTufts for technical assistance This research was supported bya Natural Sciences and Engineering Research Council ofCanada (NSERC) Discovery Grant to GRS a NationalInstitutes of Health grant to JFS (HL087216) and NationalScience Foundation grants to ZAC (IOS-1354934) and JFS(IOS-1354390)

ReferencesAudet GN Meek TH Garland T Jr Olfert IM 2011 Expression of an-

giogenic regulators and skeletal muscle capillarity in selectively bredhigh aerobic capacity mice Exp Physiol 961138ndash1150

Ayroles JF Carbone MA Stone EA Jordan KW Lyman RF Magwire MMRollmann SM Duncan LH Lawrence F Anholt RR et al 2009Systems genetics of complex traits in Drosophila melanogasterNat Genet 41299ndash307

Beaudry JL McClelland GB 2010 Thermogenesis in CD-1 mice aftercombined chronic hypoxia and cold acclimation Comp BiochemPhysiol B Biochem Mol Biol 157301ndash309

13

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

Benjamini Y Hochberg Y 1995 Controlling the false discovery rate apractical and powerful approach to multiple testing J R Stat Soc B57289ndash300

Burniston JG Meek TH Pandey SN Broitman-Maduro G Maduro MFBronikowski AM Garland T Chen Y-W 2013 Gene expression pro-filing of gastrocnemius of ldquominimusclerdquo mice Physiol Genomics 45228ndash236

Bye A Hoydal MA Catalucci D Langaas M Kemi OJ Beisvag V Koch LGBritton SL Ellingsen Oslash Wisloslashff U 2008 Gene expression profiling ofskeletal muscle in exercise-trained and sedentary rats with inbornhigh and low VO2max Physiol Genomics 35213ndash221

Cano I Mickael M Gomez-Cabrero D Tegner J Roca J Wagner PD2013 Importance of mitochondrial in maximal O2 transport andutilization a theoretical analysis Respir Physiol Neurobiol 189477ndash483

Chappell MA Snyder LRG 1984 Biochemical and physiological corre-lates of deer mouse a-chain hemoglobin polymorphisms Proc NatlAcad Sci U S A 815484ndash5488

Cheviron ZA Bachman GC Connaty AD McClelland GB Storz JF 2012Regulatory changes contribute to the adaptive enhancement ofthermogenic capacity in high-altitude deer mice Proc Natl AcadSci U S A 1098635ndash8640

Cheviron ZA Bachman GC Storz JF 2013 Contributions of phenotypicplasticity to differences in thermogenic performance between high-land and lowland deer mice J Exp Biol 2161160ndash1166

Cheviron ZA Brumfield RT 2012 Genomic insights into high-altitudeadaptation in vertebrates Heredity 108354ndash361

Cheviron ZA Connaty AD McClelland GB Storz JF 2014 Functionalgenomics of adaptation to hypoxic cold-stress in high-altitude deermice transcriptomic plasticity and thermogenic performanceEvolution 6848ndash62

Clark BC 2009 In vivo alterations in skeletal muscle form and functionafter disuse atrophy Med Sci Sports Exerc 411869ndash1875

Contractor H Stottrup NB Cunnington C Manlhiot C Diesch JOrmerod JO Jensen R Botker HE Redington A Schmidt MRet al 2013 Aldehyde dehydrogenase-2 inhibition blocks remotepreconditioning in experimental and human models Basic ResCardiol 108343

Cotton CJ Harlow HJ 2010 Avoidance of skeletal muscle atrophy inspontaneous and facultative hibernators Physiol Biochem Zool 83551ndash560

Dalziel AC Rogers SM Schulte PM 2009 Linking genotypes to pheno-types and fitness how mechanistic biology can inform molecularecology Mol Ecol 184997ndash5017

Denton JF Lugo-Martinez J Tucker AE Schrider DR Warren WC HahnMW 2014 Extensive error in the number of genes inferred fromdraft genome assemblies PLoS Comput Biol 10e1003998

Diaz I Smani T 2013 New insights into the mechanisms underlyingvascular and cardiac effects of urocortin Curr Vasc Pharmacol 11457ndash464

Duggan AT Kocha KM Monk CT Bremer K Moyes CD 2011Coordination of cytochrome c oxidase gene expression in the remo-delling of skeletal muscle J Exp Biol 2141880ndash1887

Eden E Navon R Steinfeld I Lipson D Yakhini Z 2009 GOrilla a tool fordiscovery and visualization of enriched GO terms in ranked genelists BMC Bioinformatics 1048

Egginton S 1990 Numerical and areal density estimates of fibre typecomposition in a skeletal muscle (rat extensor digitorum longus)J Anat 16873ndash80

Faulkner-Jones BE Godinho LNM Reese BE Pasquini GF Ruefli A TanSS 1999 Cloning and expression of mouse cadherin-7 a type-IIcadherin isolated from the developing eye Mol Cell Neurosci 141ndash16

Garland TJ Carter PA 1994 Evolutionary physiology Annu Rev Physiol56579ndash621

Gnaiger E 2001 Bioenergetics at low oxygen dependence of respirationand phosphorylation on oxygen and adenosine diphosphate supplyRespir Physiol 128277ndash297

Guo Y Jolly RA Halstead BW Baker TK Stutz JP Huffman M Calley JNWest A Gao H Searfoss GH et al 2007 Underlying mechanisms ofpharmacology and toxicity of a novel PPAR agonist revealed usingrodent and canine hepatocytes Toxicol Sci 96294ndash309

Gustafsson T 2011 Vascular remodelling in human skeletal muscleBiochem Soc Trans 391628ndash1632

Hayes JP 1989 Field and maximal metabolic rates of deer mice(Peromyscus maniculatus) at low and high altitudes Physiol Zool62732ndash744

Hayes JP OrsquoConnor CS 1999 Natural selection on thermogenic capacityof high-altitude deer mice Evolution 531280ndash1287

Hayward NM Yanev P Haapasalo A Miettinen R Hiltunen M Grohn OJolkkonen J 2011 Chronic hyperperfusion and angiogenesis followsubacute hypoperfusion in the thalamus of rats with focal cerebralischemia J Cereb Blood Flow Metab 311119ndash1132

Hepple RT 2000 Skeletal muscle microcirculatory adaptation to met-abolic demand Med Sci Sports Exerc 32117ndash123

Hepple RT Agey PJ Hazelwood L Szewczak JM MacMillen RE Mathieu-Costello O 1998 Increased capillarity in leg muscle of finches livingat altitude J Appl Physiol 851871ndash1876

Hochachka PW 1985 Exercise limitations at high altitude the metabolicproblem and search for its solution In Gilles R editor Circulationrespiration and metabolism Berlin (Germany) Springer-Verlagp 240ndash249

Hochachka PW 1993 Energy demandndashenergy supply coupling effi-ciency and adaptability In Bicudo JEPW editor The vertebrategas transport cascade adaptations to environment and mode oflife Boca Raton (FL) CRC Press 265-278

Hock RJ 1964 Physiological responses of deer mice to various nativealtitudesWeihe WH The physiological effects of high altitude NewYork Macmillan p 59ndash72

Jacobs RA Boushel R Wright-Paradis C Calbet JA Robach P Gnaiger ELundby C 2013 Mitochondrial function in human skeletal musclefollowing high-altitude exposure Exp Physiol 98245ndash255

Jankowski JE Londo~no GA Robinson SK Chappell MA 2012 Exploringthe role of physiology and biotic interactions in determining eleva-tional ranges of tropical animals Ecography 361ndash12

Lanza IR Sreekumaran Nair K 2010 Regulation of skeletal musclemitochondrial function genes to proteins Acta Physiol 199529ndash547

Leon-Velarde F Sanchez J Bigard AX Brunet A Lesty C Monge C 1993High altitude tissue adaptation in Andean coots capillarity fiberarea fiber type and enzymatic activities of skeletal muscle J CompPhysiol B 16352ndash58

Levett DZ Radford EJ Menassa DA Graber EF Morash AJ Hoppeler HClarke K Martin DS Ferguson-Smith AC Montgomery HE et al2012 Acclimatization of skeletal muscle mitochondria to high-alti-tude hypoxia during an ascent of Everest FASEB J 261431ndash1441

Lin J Handschin C Spiegelman BM 2005 Metabolic control through thePGC-1 family of transcription coactivators Cell Metab 1361ndash370

Lv S Cheng G Zhou Y Xu G 2013 Thymosin beta4 induces angiogen-esis through Notch signaling in endothelial cells Mol Cell Biochem381283ndash290

Ma H Guo R Yu L Zhang Y Ren J 2011 Aldehyde dehydrogenase 2(ALDH2) rescues myocardial ischaemiareperfusion injury role ofautophagy paradox and toxic aldehyde Eur Heart J 321025ndash1038

Mathieu-Costello O 2001 Muscle adaptation to altitude tissue capil-larity and capacity for aerobic metabolism High Alt Med Biol 2413ndash425

Mathieu-Costello O Agey PJ Wu L Szewczak JM MacMillen RE 1998Increased fiber capillarization in flight muscle of finch at altitudeRespir Physiol 111189ndash199

McCarthy D Chen Y Smyth G 2012 Differential expression analysis ofmultifactor RNA-Seq experiments with respect to biological varia-tion Nucleic Acids Res 404288ndash4297

McGuire JA Witt CC Remsen JV Jr Corl A Rabosky DL Altshuler DLDudley R 2014 Molecular phylogenetics and the diversification ofhummingbirds Curr Biol 24910ndash916

14

Scott et al doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from

Meyer CW Willersheuroauser M Jastroch M Rourke BC Fromme T OelkrugR Heldmaier G Klingenspor M 2010 Adaptive thermogenesis andthermal conductance in wild-type and UCP1-KO mice Am J PhysiolRegul Integr Comp Physiol 299R1396ndashR1406

Mineo PM Cassell EA Roberts ME Schaeffer PJ 2012 Chronic coldacclimation increases thermogenic capacity non-shivering thermo-genesis and muscle citrate synthase activity in both wild-type andbrown adipose tissue deficient mice Comp Biochem Physiol A 161395ndash400

Moreno-Navarrete JM Ortega F Gomez-Serrano M Garcia-Santos ERicart W Tinahones F Mingrone G Peral B Fernandez-Real JM2013 The MRC1CD68 ratio is positively associated with adiposetissue lipogenesis and with muscle mitochondrial gene expression inhumans PLoS One 8e70810

Natarajan C Hoffmann FG Lanier HC Wolf CJ Cheviron ZA SpanglerML Weber RE Fago A Storz JF 2015 Intraspecific polymorphisminterspecific divergence and the origins of function-altering muta-tions in deer mouse hemoglobin Mol Biol Evol 32978ndash997

Natarajan C Inoguchi N Weber RE Fago A Moriyama H Storz JF 2013Epistasis among adaptive mutations in deer mouse hemoglobinScience 3401324ndash1327

Oufara S Barre H Rouanet JL Chatonnet J 1987 Adaptation to extremeambient temperatures in cold-acclimated gerbils and mice Am JPhysiol Regul Integr Comp Physiol 253R39ndashR45

Ponsot E Dufour SP Zoll J Doutrelau S NrsquoGuessan B Geny B HoppelerH Lampert E Mettauer B Ventura-Clapier R et al 2006 Exercisetraining in normobaric hypoxia in endurance runners IIImprovement of mitochondrial properties in skeletal muscleJ Appl Physiol 1001249ndash1257

R Development Core Team 2014 R A language and environment forstatistical computing R Foundation for Statistical ComputingVienna (Austria) Available from httpwwwR-projectorg

Robinson M McCarthy D Smyth G 2010 edgeR a Bioconductor pack-age for differential expression analysis of digital gene expression dataBioinformatics 26139ndash140

Robinson M Oshlack A 2010 A scaling normalization method for dif-ferential expression analysis of RNA-seq data Genome Biol 11R25

Robinson M Smyth G 2007 Moderated statistical tests for assessingdifferences in tag abundance Bioinformatics 232881ndash2887

Rockman MV 2008 Reverse engineering the genotype-phenotype mapwith natural genetic variation Nature 456738ndash744

Sainson RC Harris AL 2008 Regulation of angiogenesis by homotypicand heterotypic notch signalling in endothelial cells and pericytesfrom basic research to potential therapies Angiogenesis 1141ndash51

Sanchez-Cordero V 2001 Elevation gradients of diversity for rodentsand bats in Oaxaca Mexico Global Ecol Biogeogr 1063ndash76

Schippers M-P Ramirez O Arana M Pinedo-Bernal P McClelland GB2012 Increase in carbohydrate utilization in high-altitude Andeanmice Curr Biol 222350ndash2354

Scott GR 2011 Elevated performance the unique physiology of birdsthat fly at high altitudes J Exp Biol 2142455ndash2462

Scott GR Egginton S Richards JG Milsom WK 2009 Evolution ofmuscle phenotype for extreme high altitude flight in the bar-headed goose Proc R Soc Lond B Biol Sci 2763645ndash3653

Scott GR Johnston IA 2012 Temperature during embryonic develop-ment has persistent effects on thermal acclimation capacity in zeb-rafish Proc Natl Acad Sci U S A 10914247ndash14252

Scott GR Milsom WK 2006 Flying high a theoretical analysis of thefactors limiting exercise performance in birds at altitude RespirPhysiol Neurobiol 154284ndash301

Scott GR Richards JG Milsom WK 2009 Control of respiration inflight muscle from the high-altitude bar-headed goose and low-al-titude birds Am J Physiol Regul Integr Comp Physiol 297R1066ndashR1074

Sheafor BA 2003 Metabolic enzyme activities across an altitudinal gra-dient an examination of pikas (genus Ochotona) J Exp Biol 2061241ndash1249

Stager M Swanson D Cheviron ZA 2015 Regulatory mechanisms ofmetabolic flexibility in the Dark-eyed Junco (Junco hyemalis) J ExpBiol 218767ndash777

Stone EA Ayroles JF 2009 Modulated modularity clustering as a tool forfunctional genomic inference PLoS Genet 5e1000479

Storz JF Natarajan C Cheviron ZA Hoffmann FG Kelly JK 2012Altitudinal variation at duplicated beta-globin genes in deer miceeffects of selection recombination and gene conversion Genetics190203ndash216

Storz JF Runck AM Moriyama H Weber RE Fago A 2010 Geneticdifferences in hemoglobin function between highland and lowlanddeer mice J Exp Biol 2132565ndash2574

Storz JF Runck AM Sabatino SJ Kelly JK Ferrand N Moriyama H WeberRE Fago A 2009 Evolutionary and functional insights into themechanism underlying high-altitude adaptation of deer mousehemoglobin Proc Natl Acad Sci U S A 10614450ndash14455

Storz JF Scott GR Cheviron ZA 2010 Phenotypic plasticity and geneticadaptation to high-altitude hypoxia in vertebrates J Exp Biol 2134125ndash4136

Suarez RK Moyes CD 2012 Metabolism in the age of ldquoomesrdquo J Exp Biol215 2351ndash2357

Supek F Bosnjak M Skunca N Sumc T 2011 REVIGO summaries andvisualizes long lists of Gene Ontology terms PLos One 6e21800

Templeman NM Schutz H Garland TJ McClelland GB 2012 Do micebred selectively for high locomotor activity have a greater relianceon lipids to power submaximal aerobic exercise Am J Physiol RegulIntegr Comp Physiol 303R101ndashR111

Van Sant MJ Hammond KA 2008 Contribution of shivering andnonshivering thermogenesis to thermogenic capacity for thedeer mouse (Peromyscus maniculatus) Physiol Biochem Zool 81605ndash611

Ventura-Clapier R Kuznetsov A Veksler V Boehm E Anflous K 1998Functional coupling of creatine kinases in muscles species and tissuespecificity Mol Cell Biochem 184231ndash247

Wagner PD 1996 Determinants of maximal oxygen transport and uti-lization Annu Rev Physiol 5821ndash50

Wallimann T Wyss M Brdiczka D Nicolay K Eppenberger HM 1992Intracellular compartmentation structure and function of creatinekinase isoenzymes in tissues with high and fluctuating energy de-mands the ldquophosphocreatine circuitrdquo for cellular energy homeosta-sis Biochem J 281(Pt 1)21ndash40

Walsh B Hooks RB Hornyak JE Koch LG Britton SL Hogan MC 2006Enhanced mitochondrial sensitivity to creatine in rats bred for highaerobic capacity J Appl Physiol 1001765ndash1769

Weibel ER 1979 Stereological methods Toronto (ON) Academic PressWilliams CK Li J-L Murga M Harris AL Tosato G 2005 Up-regulation of

the Notch ligand Delta-like 4 inhibits VEGF-induced endothelial cellfunction Blood 107931ndash939

Zhang B Zhang Y La Cour KH Richmond KL Wang XM Ren J 2013Mitochondrial aldehyde dehydrogenase obliterates endoplasmic re-ticulum stress-induced cardiac contractile dysfunction via correc-tion of autophagy Biochim Biophys Acta 1832574ndash584

15

High-Altitude Adaptation in Deer Mice doi101093molbevmsv076 MBE at U

niversity of Nebraska-L

incoln Libraries on A

pril 20 2015httpm

beoxfordjournalsorgD

ownloaded from